In order to record the contents of a document with ease, without using a device like a scanner, portable image capturing devices like digital cameras are sometimes used for capturing and recording an image of such a document. When a document is shot for such a purpose, it is ideal that the document be shot right from the front so that characters on the document will not be captured to be obscure or unreadable. However, there are cases where a person shooting a document can not stand up (during a meeting, etc.) and has to shoot the document on a desk from an oblique direction.
Japanese Patent Provisional Publication No. 2002-334327 (hereinafter referred to as a “document No. 1”) describes an image processing device which converts an inputted image (in which the subject (e.g. a document) has been captured from an oblique direction and characters which are blurry or slanted are hard to read) into an image in which the subject captured in the inputted image is viewed from a different viewpoint allowing easy reading of the characters.
The image processing device described in the document No. 1 analyzes the image in which the document has been captured, recognizes line bundles (arrangement of characters, edges of the document, etc.), and judges the rotation angle of the document image in the whole image based on the line bundles. Further, the image processing device converts the image of the document into an output image in which the document has been rotated to a direction in which character strings in the document are arranged horizontally (such a direction will hereinafter be referred to as an “erect direction”) and the document is viewed from the front.